Office Soon | free documents download

The Private Rate of Return to Schooling: Evidence from Eritrea Temesgen ...

Essays in Education Volume 21, Summer 200777 The Private Rate of Return to Schooling: Evidence from Eritrea Temesgen Kifle The University of Queensland Abstract
The extent of links between education and earnings is a determining factor in making
decisions about investment in education. The purpose of this study is to estimate the
private rate of return to education in Eritrea using sample data from employees working
in public and private sector of the economy. The main result obtained with the help of
extended Mincerian earnings function indicates the financially rewarding effect of
education. It is also found that the rates of return to education increase with the increase
in levels of education. The findings imply a need for expanding access to education and
the possibility of sharing the cost burden of education, especially at tertiary level of
education. Introduction Education is often considered as the single most important determinant of a person’s economic and social achievement. Education provides both direct and indirect
benefits for the individual who receives the education and the society with which this
individual connects. At a national level, the effect of investment in human capital on
productivity, technology and growth has long been stressed by economists. For poor
countries like Eritrea education plays a key role in poverty alleviation. Recognizing these benefits, the government of Eritrea has been making an effort to increase access to education, thereby improving children’s school participation.
Between the academic year 1990/91 and 2001/02, the number of elementary, middle and
secondary schools owned by the government increased 4.5, 3.5 and 2.3 times
respectively. Within this period of time, the number of students enrolled in government
schools increased 4.0 times in elementary, 3.5 times in middle and 2.1 times in secondary
level of education. This is good progress considering that Eritrea is a newly independent
country. However, research on the economic benefit that arises from acquiring education
is non-existent in the country. There is no documented information on the incentives for
human capital accumulation. There is no research on whether acquisition of education
provides individuals enough incentive to invest. The focus of this paper is on private returns to education for formal sector workers only and thus it should be interpreted as conditional on having a wage/salary
earnings job in the formal sector of the economy. Empirical evidences of the return to
investment in education can be used as a reliable guide to design educational policy in Essays in Education Volume 21, Summer 200778Eritrea, mainly concerning the efficient allocation of scare resources between the
different levels of education. The structure of the paper is as follows. Section two provides review of literature on the economic theory of education and earnings. Section three provides basic facts
about education in Eritrea. Section four focuses on method of estimating the model. The
database is described in section five. Section six offers the estimated results of regression
analysis. Finally, section seven presents conclusion, implication and recommendation. Review of Literature Returns to years of education in different countries are heavily influenced by the supply of workers with different amount of education (Schultz, 1999). In turn the supply
of workers with different levels of education is influenced by government policy choices.
The concept of human capital refers to the fact that human beings invest in themselves to
raise their future income by increasing their lifetime earnings. Given that people freely
choose to invest, perfect capital markets exist and mobility of labour is not hindered, the
human capital framework assumes that wage differentials reflect returns to investment in
human capital (Terrell, 1989). Though it might not be true for each and every individual
and the correlation is far from perfect, it is true for the average person that the amount of
education an individual possesses is positively correlated with personal earnings. The
first model of this kind that was developed by Mincer (1958, 1970, 1974) assumes a
complete absence of environmental inequalities and takes the length of schooling as a
basic source of heterogeneity of labour incomes. The theory assumes that in the absence of serious market imperfections earnings are equivalent to worker’s marginal product and that the greater the worker’s stock of
human capital the greater is his/her productivity and hence earnings (Shah, 1986). There are several estimates of the returns to schooling for wage workers that do not consider the selection bias issue as a serious limitation (van der Gaag & Vijverberg,
1989; Griffin & Edwards, 1993; Vieira, 1999; Duraisamy, 2002; Trostel, Walker, &
Woolley, 2002; Bjorklund & Kjellstrom, 2002; Li, 2003; Moock, Patrinos, &
Venkataraman, 2003; Hawley, 2004; Martins & Pereira, 2004; Silles, 2005; Yang, 2005).
Reasons for excluding the self-employed or unpaid family workers are non-availability of
data, which is the case in this paper, low proportion of employees outside the wage
employment, the presumption that a failure to adjust for selection bias has little or no
impact on estimates of schooling returns, as written by Dearden (1998), and intentional
focus on the wage labour force, particularly on the formal sector employees. Based on a Mincerian earnings function method a number of studies on education and earnings in Africa found that the private rate of return to an additional year of
schooling was quite high. For countries in Africa it is commonly asserted that the private
returns to investment in education are highest at primary level and thus primary education
should be the number one investment priority (Psacharopoulos, 1985, 1994). However, a
number of recent studies on education in Africa have found that the private rates of return Essays in Education Volume 21, Summer 200779not only are relatively lower than suggested in the conventional pattern but also increase
with the level of education. Challenging the view expressed by Psacharopoulos, Bennell
(1996) argues that the conventional rate of return on education patterns almost certainly
do not prevail in Sub-Saharan Africa (SSA) under current labor market conditions.
Surveys of estimates for Botswana by Siphambe (2000), for Cameroon, Ghana, Kenya,
Zambia and Zimbabwe by Bigsten, Isaksson, Soderbom, Collier, Zeufack, Dercon et al.
(2000), for South Africa by Mwabu & Schultz (2000), for Ghana by Jones (2001), for
Kenya by Wambugu (2001), for Nigeria by Armolaran (2002), for Burkina Faso, Cote
d’Ivoire, Ghana, Kenya, Nigeria and South Africa by Schultz (2003), for Kenya and
Tanzania by Soderbom, Teal, Wambugu, & Kahyarara (2004), for Burkina Faso by
Kazianga (2004), for Cameroon by Amin & Awung (2005), and for Rwanda by Lassibille
& Tan (2005) show that returns to education rise with educational attainment. The links between education and earnings are of deciding factor to decisions about the efficient allocation of resources. However, due to omitted variables,
interpretation of such estimates is usually qualified by comment on possible upward
biases. Behrman & Deolalikar’s (1993) criticism is that the studies which typically
attribute the association between years of schooling and wages do not include a host of
other factors that plausibly may be correlated with years of schooling that affect wages. Moll (1998) stressed the point that years of schooling, as an input measure of human capital, may influence the wage if it captures other elements. Topel (1991) has
concluded that, other things remaining constant, 10 years of job tenure raise the wage of
the typical worker by over 25%. The strong positive relationship between tenure and
wage rates was also assessed by Altonji & Williams (1997). The strong long term
employer-employee relationship conditioned by promotion provisions was mentioned by
Theodossiou (1996) to specify the significant effect of tenure on wages. Firms, in order
to discourage labour turnover and inter-firm mobility, establish long-term employment
relationships with their most highly valued employees. Thus, employees with longer
tenure with their current employer have higher earnings than other employees with the
same total work experience but relatively shorter tenure. Opposing the significant effect of tenure on wages, Altonji & Shakotko (1987) argued that the partial effect of tenure on wages was small because the strong relationship
between tenure and wages was due primarily to heterogeneity bias across individuals and
across job matches. Similarly, Jacobson, Lalonde, & Sullivan (1993) have found that high
tenure workers separating from distressed firms suffer long term losses averaging 25%
per year. Re-examining the wage-tenure relationship, Williams (1991) has found that
tenure increases wages only in the first several years of employment. The occupation in which a worker is employed has an important effect on the level of his/her wages and salaries. Disparities in earnings between different occupations
have been often noticed in less developed countries than in developed countries (Kothari,
1970). Earnings differentials would not indicate compensating differentials but rather
signal enlarged inequalities because some individuals not only are denied the possibility
of working at high and satisfied job levels but also have to accept lower wages (Hartog, Essays in Education Volume 21, Summer 2007801986). For that reason the reward for education differs substantially by the job level at
which an individual is occupied. The argument against the above assertion is that
occupation and jobs are irrelevant entities in explaining earnings differentials because
market forces tend to equate rates of return throughout and thus equilibrium situation will
exist in the long-run. It is widely accepted that family background affects education by influencing the amount of education individuals obtain. Family background and influences are more
important in determining education and earnings (Krishnan, 1996). Altonji & Dunn
(1996) explored the possibility that the education slopes of wage equations are influenced
by family background as measured by father’s and mother’s education. Beach & Finne
(1988) have also researched the positive effects of parents’ education on son’s
educational attainment, which this ends-up increasing importance of indirect and total
effects of the family background variables on earnings. Controlling the workers’ own
schooling and the schooling of other relatives, Lam & Schoeni (1993) have discovered
the relationship between having a father with a university education and getting a 20%
wage advantage when compared with illiterate father, and a 9% wage advantage when
compared to a father with 4 years of schooling. Sahn & Alderman (1988) have pointed
out that the wage offer in developing countries is influenced by other genetic and
environmental influences captured in the wage of one’s father. Thus, the significant
impact of family background on earnings could mean that family background determines
the quality of education and learning environment at home (as educated parents can
improve the educational opportunities of their children through their absorption of
attitudes and acquisition of human capital) or would indicate that individuals from a
better family background are able to get the better jobs through family connections and
influences. The partial cause of earnings differentials may also be sector of employment. Mann & Kapoor (1988) have explored that, on the average, public sector workers are
paid much higher wages than the private and joint sector workers. Rees & Shah (1995)
have reasoned that the private wage determination is subject to profit constraint, whereas
the public sector wage determination is subject to an ultimate political constraint. Thus,
wages in the public sector are higher than in the private sector. Pritchett (1999)
highlighted the situation in which governments are taking resources away from non-
governmental activity in the form of taxes so as to pay additional workers whose
marginal product in the public sector is very low but are paid much higher wages than
workers in the private sector. On the whole, even if it is widely accepted that level of schooling is positively correlated with level of earnings, it is not acceptable to understate the effects of other
influential factors on earnings. This implies that the basic Mincerian approach to estimate
return to education should be extended. Essays in Education Volume 21, Summer 200781Education in Eritrea Before the Italian domination, education in Eritrea had been divided into two broad categories, namely local and religious. While local education was composed of
training children in practical and productive skills, religious education for small numbers
of children was provided by Christian and Muslim clerical hierarchies. During the Italian colonial period (1890-1941), education in Eritrea was not widely available for Eritreans. Even if for Italians living in Eritrea schooling was
compulsory to age 16, the highest level allowed for an Eritrean to reach was only fourth
grade. After almost sixty years of Italian colonization in Eritrea only a small,
predominately male segment of the population could claim rudimentary schooling
(Stefanos, 1997). Unlike Italians, the British were less strict about educational opportunities. Under the British Administration (1941-1952) Eritreans were allowed to work in lower
administrative positions and to form political parties and trade unions, and thus the
number of elementary schools was increased and new middle and secondary schools were
opened (Rena, 2003). During this time, not only education was expanded in villages and
towns but also textbooks, in both local and foreign languages, became available. As a
result, the desire for education increased very rapidly. Educational facilities in Eritrea were far better than in Ethiopia at the time when Eritrea passed from British Administration to the federal arrangement in 1953. However,
the Ethiopian government began to weaken Eritrean education. Eritrean languages were
banned and substituted by the Ethiopian language and all education decisions began to be
made in the capital of Ethiopia (Gottesman, 2002). Besides, remuneration to Ethiopian
teachers was more than Eritrean counterparts. The strikes by Eritrean teachers for more
pay didn’t bring about a positive change; to the contrary, a number of Eritrean teachers
were arrested or transferred to Ethiopia (Gottesman, 2002). In consequence of this, the
quality of education in Eritrea further deteriorated. During the Ethiopian socialist regime (1974-1991) educated Eritreans were a particular target of harassment and violence. Intolerably, many Eritrean teachers and
students began to join the armed struggle for independence; and this made worse the
already deteriorated quality of education in Eritrea. During the war, the Eritrean People’s
Liberation Front (EPLF), by establishing educational programs, facilitated the provision
of education in regions under its control. Education was seen by EPLF leaders as integral
to the national liberation struggle. Educational efforts, which primarily focused on the
freedom fighters, diversified with the provision of education to orphans, refugees,
children of fighters and those who had run away to join the front but were too young to
fight. In addition to this, adult literacy campaign and regular schools were maintained in
liberated areas. After the 1991 liberation and the 1993 vote for independence, a macro policy document that incorporated educational policy was designed by the Government of the Essays in Education Volume 21, Summer 200782State of Eritrea (GSE). Eritrea’s main educational goals for the next years will emphasize
the importance of early childhood education care and development, formal basic
education, and adult literacy (MOE, 1999). To achieve these general objectives and
feature goals, various policies have been designed to: make universal primary education
(up to seven years) available to all; increase enrolments at the secondary, technical and
vocational schools (to meet skilled manpower requirement of both public and private
sector); promote education through formal and informal channels (to achieve higher
literacy rate and enhance competence); expand selective tertiary education through the
utilization of training opportunities (to meet manpower requirement of the country);
expand technical and vocational training (to enhance the job adaptability and retaining
potential of the students); promote equal educational opportunity in terms of access,
equity, relevance and continuity of education; make elementary education accessible to
children in their mother tongue; promote and encourage pre-primary education in all
zones; provide a minimum of three years of adult education; encourage education through
private sector; and to provide middle and secondary school in English (GSE, 1994; MOE,
2000a). In adopting these policies, the government, the community and the direct
beneficiaries will be made to contribute varying amounts towards financing costs of
education. In Eritrea, schools are administered either by the Ministry of Education (MOE) or by other institutions (MOE, 2000b). Those schools that are administered by the MOE are
called Government schools, whereas those schools that are managed by institutions other
than the MOE are called Non-government schools. Educational provision in Eritrea
begins at pre-school level and ends at tertiary education. Children in the age group 5-6
are expected to enroll in pre-primary school and the length of time in this level of
education is two years. The elementary level of education covers 5 years and a theoretical
age range of 7 to 11. As a second part of compulsory schooling, the middle level takes
two years (grade 6 and 7) to complete and includes students within a theoretical age
category of 12 to 13. In this level of education, the medium of instruction is English. The
time during which secondary level lasts is four years (8-11) and a theoretical age range of
14-17. Similar to middle level, the medium of instruction at this level of education is
English. Since independence education in Eritrea has expanded enormously. Between 1990/91 and 2001/02, the number of schools in elementary, middle and secondary levels
increased by around 225%, 158%, and 132% respectively, and the number of students
enrolled increased by around 203%, 194%, and 118% respectively (see Table 1). Though
it is still low in comparison with other low income countries, Eritrea reached 66% Gross
Enrolment Ratio (GER) in primary level of education in 2004 (see Table 2). But, despite
improvements, access to education is still limited. There exists gender and regional
disparities in education. Essays in Education Volume 21, Summer 200783Table 1
Eritrea: Schools and enrolment by level of education and ownership 1990/91a2001/02 SchoolsEnrolment Schools Enrolment
LevelTotal GovbTotalGovTotalGovTotalGovElementary214137109087 74878695610330278303009Middle593727556216711521308088275124Secondary1916321413151944367018367511a1990/91 indicates the last pre-liberation academic year in Eritrea.bGov is an abbreviation for government.Source: MOE, 2004
Table 2
Gross Enrolment Ration (GER)a by level of educationLevel Eritrea Low-income countries Pre-primary 7 27 Primaryb 66 100 Secondary 28 46 Tertiary 1 9 aGER is the ratio of total enrolment, regardless of age, to the population of the age group that officiallycorresponds to the level of education shown.bPrimary level covers grade 1 to 7.Source: World Bank, 2006 Empirical Model The specification of the econometric model of the private rates of return to education can be based on economic theory and on any available information relating to
the phenomenon being studied. Hence, on the basis of information on the economic
theory of education and earnings, we may write the basic earnings function in the general
form: E = (S, EX), where E is wage and salary earnings, S indicates years of schooling and EX is number of
years of work experience. The basic earnings function method is due to Mincer and involves the fitting of a semi-log ordinary least squares regression using the natural logarithmic of earnings as a
dependent variable and years of schooling and potential years of labour market
experience and its square as explanatory variables. In an equation form: uEXSQbEXbSbbE++++=3210 ln, where Eln stands for logarithm of monthly earnings, S for years of educational attainment, EX for number of years of work experience, and ufor measurement error. In this equation, number of years of work experience is squared )(EXSQto capture the declining effects of experience as individual ages. The above model has been widely criticized for its shortcomings. First, using this method one cannot estimate returns to education at different levels because the Essays in Education Volume 21, Summer 200784coefficient on years of schooling can only be interpreted as the average private rate of
return to one additional year of education, regardless of the educational level in which
this year of school refers to. Second, a person’s earnings can also be determined by
factors other than years of schooling and years of labour market experience. Because of
these points, the extended earnings function method, which converts the continuous years
of schooling variables into a series of dummy variables and includes other additional
variables, is found to be the best way of estimating returns to education. Thus, the
earnings function becomes: ),, (ln =niX,EX, EXSQEDE where iED is a series of educational dummy variables (that allows for wage differences among four levels of education) and nXis a vector of other wage determining characteristics. The estimated regression equation for the extended version of Mincer’s
model can be stated as: u,SE bGE bME b FE bTE bMHW bOCCbEXSQbEXbEDbbEii+++++++++++=109876543210lnln whereiOCC is a dummy variable indicating occupation, MHWlnis logarithm of hours worked per month,TE is number of years of tenure in the current job, FE is father’s level
of education, ME is mother’s level of education, GE is a dummy variable indicating
gender and SE is a dummy variable signifying sector of employment.1 The above earnings equation is developed in the context of wage and salary income in the formal sector of the economy and thus the returns to education estimated
here are for formal sector workers only and give no explanation of the rates of return in
other sectors. Data and Descriptive Analysis The data is drawn from 363 employees (salary and/or wage earners) working in public and private sectors of the Eritrean economy. The sample does not include the self-
employed because (besides lack of data) it is difficult to separate their wages from profit
income.2 In addition, unregulated wage workers are excluded, since data were not available.3 Data collection was carried out by the author during 2001-2002. The population was first divided into two groups, namely public and private sector
employees; then a sample of 212 public sector employees and a sample of 151 private
sector employees were drawn randomly and proportionately. Again, a simultaneous
proportional stratification for a variable gender was done with separate simple random
sample. Out of the total respondents, 184 were female. 1 Due to shortage of data the model does not control for all explanatory variables that affect individual’searnings.2 It is known that earnings data for self-employed also include returns to physical capital and to risk anduncertainty bearing, which are difficult to disentangle from returns to human capital.3 The exclusion of data on employees working in the informal sector cannot significantly alter the results,as it is almost impossible in Eritrea to run a business without having a license.Essays in Education Volume 21, Summer 200785 In this study, the dummy explanatory variable for education (iED ) is categorized as primary (grade 1 to 7), secondary (8-12), post secondary non tertiary (13-14) and
tertiary (15+). The mnemonic names that are given for primary, secondary, post
secondary non tertiary, and tertiary level are ED1, ED2, ED3, and ED4 respectively.
Concerning occupation dummy (iOCC ), it is difficult to classify respondents according to the International Standard Classification of Occupations (ISCO) because most of the
respondents did not know clearly what their occupation called.4 Some of them perform different tasks, so it was not easy to find one occupational group that fits the ISCO. Thus,
a summary of five occupational categories has been made and classified as OCC1,
OCC2, OCC3, OCC4, and OCC5. OCC1 stands for legislators, senior officials, managers
and professionals; OCC2 is a symbol for technical and associate professionals; OCC3 is
an abbreviation for clerks; OCC4 stands for service workers, agricultural and fishery
workers and those who do elementary occupations; and OCC5 is a symbol for craft and
related trade workers and plant and machinery operators. 4 The ISCO refers to the classification of the International Labour Organisation (ILO).Essays in Education Volume 21, Summer 200786
Table 3
Descriptions, means and standard deviation of variables VariablesDescriptionMeanS. D.EMonthly earnings (in Nakfa )a981.3146 601.44455SYears of schooling10.86093.92654ED1Primary level, 1 if the employee attained primary level, 0
otherwise
0.2534
0.43558ED2Secondary level, 1 if the employee attained secondary level,
0 otherwise
0.3912
0.48869ED3Post secondary non tertiary level, 1 if the employee attained
post secondary non tertiary level, 0 otherwise
0.2259
0.41875ED4Tertiary level, 1 if the employee attained tertiary level, 0
otherwise
0.1295
0.33619EXYears of work experience14.377811.56204MOHMonthly hours worked191.8579 37.17504TENumber of years of tenure8.48369.39882FEFather’s level of education3.97524.59365MEMother’s level of education1.62533.32455SESector of employment, 1 if the employee works in public
sector, 0 otherwise
0.5840
0.49357GEGender, 1 if the employee is male, 0 otherwise0.49310.50064OCC1Legislators, senior officials, managers and professionals, 1 if
the employee belongs to OCC1, 0 otherwise
0.1433
0.35081OCC2Technical and associate professionals, 1 if the employee
belongs to OCC2, 0 otherwise
0.2039
0.40342OCC3Clerks, 1 if the employee is a clerk, 0 otherwise0.31130.46366OCC4Service workers, agricultural and fishery workers and those
who do elementary occupations, 1 if the employee belongs
to OCC4, 0 otherwise
0.2176
0.41320OCC5Craft and related trade workers and plant and machinery
operators, 1 if the employee belongs to OCC5, 0 otherwise
0.1240
0.330aNakfa is the standard unit of money in Eritrea. In 2001, 1 Nakfa = 0.0738 USD.
As can be noted from Table 3, the mean of schooling of the respondents was 10.9, and
their average monthly salary was 981.3 Nakfa. The average of years of work experience
and job tenure was 14.4 and 8.5 respectively. If classified by level of education, 25.3% of
the respondents had some primary education; 39.1% had reached secondary level; 22.6%
had post secondary non tertiary education; and 13.0% had tertiary education. On average,
employees worked 191.8 hours per month.5 By type of occupation, 14.3% of the interviewees were OCC1; 20.4% were OCC2; 31.1% were OCC3; 21.8% were OCC4;
and the rest 12.4% were OCC5. On average, the number of years of schooling parents
completed was around 4 years for the father and 1.6 years for the mother. As can be seen in Table 4, there is a positive relationship between level of education and average monthly earnings. For instance, on average those with some 5 The standard workweek is 44.5 hours in Eritrea.Essays in Education Volume 21, Summer 200787secondary education earn 1.46 times than those who have some primary education. When
one level of education is compared with its subsequent level, the highest wage difference
is found between those with tertiary education and those with post secondary non tertiary
education.
Table 4
Mean monthly earnings of workers by level of education OverallEDEDi+1iED1576.5 -ED2840.11.46ED31189.51.41ED41881.21.58Total981.3 -Source: Author’s calculation. Results of Regression Analysis In Table 5, the regression results are presented in three columns. While the first column shows the estimates from the basic Mincerian earnings function, the second
column presents the results of the extended model. The third column, which includes a
quadratic term in schooling, shows that earnings are convex in education. In column 1, an
additional year of schooling increases monthly earnings by 11%. Similarly, one more
year of work experience adds 2% to earnings (but at a decreasing rate -0.000119). The
negative sign of the variable EXSQ indicates that the impact of experience on earnings is
like a hill-shaped parabola. In studies of earnings functions, it is suggested that the impact
of experience diminishes as the amount of experience increases. This is based on notions
of physical and mental ageing, diminishing returns, and optimal investment in human
capital. In this model, the proportion of the variation in earnings which is explained by
years of schooling and work experience is 55.8%. The F-test of goodness of fit of
regression shows that the explanatory variables do actually have a highly significant
influence on earnings. The education dummy variables in column 2 show the differential earnings for an individual with primary or secondary or post secondary non tertiary education relative to
an individual with tertiary education. Accordingly, the earnings differential for an
individual with primary education relative to an individual with tertiary education is -
55.4%, whereas for secondary and post secondary non tertiary the earnings differentials
relative to tertiary education are -46.1% and -28% respectively. These coefficients show
that the rates of return to education increase with the increase in levels of education. The
first year of work experience is worth 2.5%. Holding other explanatory variables
constant, the peak wage occurs after about 38 years of work experience. When the
coefficient on EX is positive and the coefficient on EXSQ is negative, the quadratic has a
parabolic shape. There is always a positive value of EX, where the effect of EX on lnE is
zero. Before this point (the point that makes the effect of EX on lnE zero) EX has a
positive effect on lnE; after this point EX has a negative effect on lnE. This point is Essays in Education Volume 21, Summer 200788known as the turning point. To know the turning point we have to divide the coefficient
on EX by 2 times the absolute coefficient on EXSQ. From column 2 of Table 5, we find
that the turning point for EX is 37.95 years.6 This implies that the return to EX becomes zero at around 38 years. This is due to the fact that few employees in the sample (only
3.7% of the interviewees) had more than 38 years of experience, and so the part of the
curve to the right of 38 can be ignored.7
Table 5
Earnings regressiona Variableb (1) (2)(3)Constant5.278 (73.115)6.642 (15.697)6.156 (14.134)S0.110 (19.845)***-0.039 (-1.757)*SSQ0.004 (3.736)***ED1-0.554 (-5.509)***ED2-0.461 (-5.354)***ED3-0.280 (-3.315)***EX0.02037 (3.365)***0.025 (4.373)***0.023 (4.028)***EXSQ-0.000119 (-0.770)0.0003 (-2.476)**0.0003 (-2.095)**lnMOH0.114 (1.437)0.103 (1.304)TE-0.004 (-1.832)*-0.005 (1.928)*FE0.012 (2.390)**0.013 (2.537)**ME-0.009 (-1.220)-0.009 (-1.248)GE0.099 (2.383)**0.114 (2.744)***SE-0.047 (-1.154)-0.046 (-1.152)OCC2-0.143 (-1.765)*-0.131 (-1.487)*OCC3-0.366 (-4.401)***-0.384 (-4.936)***OCC4-0.924 (-9.503)***-0.868 (-8.68)***OCC5-0.463 (-4.794)***-0.417 (-4.274)***R2(adjusted)0.5580.7010.705F-value153.255***57.513***62.487***Durbin-Watson 1.7081.7361.761Sample size363363363aValues in parentheses are t-statistics. Asterisks indicate level of significance; *, significant at 10 percent;**, significant at 5 percent; ***, significant at 1 percent.bIn our model, ED4 and OCC1 are the omitted reference groups.
The coefficient on TE in column 2 paradoxically shows a negative sign, which this
implies a negative relationship between earnings and tenure. This can be an indication of
the advantage of labour mobility over job tenure. According to Ransom (1993) the reason
why employees stay more years in their current job despite the advantage that could be
got from job mobility is because mobility cost is high or employers have monopsony
power. In the case Eritrea, however, it is difficult for public servants, especially higher- 6=95.3706)x(-0.0003320250906.07 For more information refer to Mirer (1983) and Wooldridge (2003). Essays in Education Volume 21, Summer 200789skilled employees, to leave their jobs after having accepted civil service employment
(UNIDO, 2003). The effect of father’s education on employee’s earnings is around 1.2%,
and this implies that father’s educational achievement is important in motivating children
to invest in education and hence higher future earnings. The coefficient on variable GE
indicates that, for the same levels of explanatory variables inserted into the model, male
employees earn about 10% more than female ones. The dummy coefficients on
occupation show that earnings in OCC2, OCC3, OCC4 and OCC5 are lower by 14.3%,
36.6%, 92.4% and 46.3% respectively when compared with earnings in OCC1. While ED4 in our model is assumed as reference group, the rate of return to the K-1 level of education )rk-1( is estimated by subtracting the coefficient of KEDfrom that of 1k-ED and divided by the average number of years of schooling (AS) at the K level.8 For instance, to calculate the annual private rate of return to schooling for ED1 (versus
ED2) based on the estimated coefficients column 2 of Table 5, kkk-k-ASEDEDr =11, where 1k-r is the rate of return to ED1 level of education, 1k-ED is estimated coefficient of ED1, kED is estimated coefficient of ED2 and kAS is average years of schooling at ED2 level. 1.41%-or 01411.04.25(-0.37)-0.43-1==k-r In this fraction, numbers above the line (i.e. coefficients of ED1 and ED2) are adjusted
by[]exp(b ) -1i.9 Using the above calculation procedure, the annual private rate of return to schooling for ED2 (versus ED3) is less by 9.09%, and for ED3 (versus ED4) it is less
by 10.43%. Even if the data limit us from analysing the annual private rate of return of those with no education (versus primary level), it seems that the highest annual private rate of
return is in ED4. The above estimates imply that the annual private rates of return to
different levels of schooling are increasing. In ED4, the rate of employment creation is
adequate to absorb all the graduates entering the labour market. One can say that primary 8 For further understanding see the writings of Halvorsen & Palmquist (1980) and Siphambe (2000). In oursample study, the average number of years of schooling for ED1, ED2, ED3 and ED4 is 5.20, 4.25, 1.43,
and 2.30 respectively.9 The straight interpretation of slope coefficients in log-linear and semi-log models is appropriate onlywhen a change in explanatory variables and the accompanying change in dependent variable are small. For
large changes, it is advisable to interpret after correcting the estimates using the formulas[]exp(b .ln2) -1i for log-linear and[]exp(b ) -1ifor semi-log (The number “exp” is the base of naturallogarithms and it is approximately equal to 2.7183. “ln” stands for logarithm andbi for parametersassociated with respective variables). This implies that the former formula has an effect of reducing the
values of both negative and positive coefficients, whereas the latter has an effect of reducing the values of
negative coefficients and increase the values of positive coefficients when compared with the original
estimates. (For more information refer to Mirer (1983).Essays in Education Volume 21, Summer 200790education is the most affected because the earnings differentials between employees with
primary education and those with secondary education are very small. Since monetary
losses of not going to secondary level are relatively very small (1.41%) those whose level
of education is primary are financially less motivated to achieve secondary level and thus
they are the most affected. In the above calculation, one can observe that the direct costs of education are not taken into account. The standard earnings equation (that was developed by Mincer)
assumes that earnings foregone while attending school are the only costs of education. It
assumes no direct costs of education. It is generally known that the direct costs of
education are insignificant when compared with indirect costs, but it is not reasonable to
assume that all foregone earnings are obvious. One should not forget that even a person
who invested in education could be unemployed. Therefore, the calculated results are
only applicable when one assumes that the direct costs of education are nominal and
those who decide to proceed with their education would certainly lose income from
employment. The estimated results of one level of education versus the other can be confirmed when a quadratic term in schooling (SSQ) is inserted into the model to allow for a
systematic change in schooling coefficient with changing levels of schooling. According
to Mincer (1974), a significant negative coefficient at SSQ implies that rates of return are
lower at higher levels of education. In our case, the coefficient on SSQ is highly
significant and positive, which this implies higher rates of return with higher levels of
schooling (see column 3 of Table 5). When the coefficient on S is negative and the
coefficient on SSQ is positive, the quadratic has a convex shape. To know the turning
point for the education variable we have to divide the coefficient on S by 2 times the
absolute coefficient on SSQ, and the approximate result is equal to 4.5. In our data, about
8.5% of the interviewees in the sample had less than 5 years of schooling, and this is not
so small a percentage to ignore. However, it is also difficult to believe that starting at
three and increasing to four years of schooling actually reduces the return to education.
One possible explanation is that few employees in the sample, despite having less years
of schooling than the turning point, had extremely high earnings associated with long-
term work experience. Another explanation is that our model does not capture all
independent variables.10 The evidence that earnings are convex in education implies that the assumptions made by Mincer regarding linearity in schooling and separability
between education and experience cannot be applied. The findings of this paper are consistent with the estimated results for most SSA countries (see Tables 6 and 7). Two decades ago, it had been asserted that returns to
education in developing countries decreased with the increase in levels of schooling and
thus governments of these countries should heavily invest in primary education. But,
recent studies have revealed that the completion of primary level does not guarantee
higher earnings, as many countries in SSA have already made great progress in
promoting primaryeducation. Since returns to education are highest at higher education 10 For more information on possible explanations refer to Wooldridge (2003)Essays in Education Volume 21, Summer 200791this level of education should also be expanded and students at this level should pay
tuition fees so as to offset part of the huge government expenditure on tertiary education.
Table 6
Spline estimates of wage return per year of schooling in selected SSA countriesCountryYearCoefficient (in %)SourceBotswana1979
1993-199419.1
12.0Lukas & Stark (1985)
Siphambe (2000)Cote d’Ivoire198620.1van der Gaag & Vijverberg (1989)Eritrea1996-19978.2, for male heads
10.4, for female headsArneberg (1999)Ethiopia19728.0Psacharopoulos (1985)Ghana19898.5Glewwe (1991)Kenya197016.4Psacharopoulos (1985)Rwanda1999-200117.5Lassibille & Tan (2005)Tanzania198011.9Psacharopoulos (1985)
Table 7
Implied private rate of return in percent per annum, selected SSA countries BotswanaaCameroonbCote d’IvoirecGhanaKenyadRwandaeYear1993-199419941985-1987199819941999-2001Primary vs. None75.2 (Pooled)
11.8 (Male)
1.9 (Female)15.0 (Male)
4.5 (Female)
17.0 (Male)
10.0 (Female)8.4 (Male)
2.5 (Female)19.4Middle vs. Primary83 ---14.0 (Male)
9.6 (Female)
12.0 (Male)
3.8 (Female)4.6 (Male)
4.9 (Female)11.0 (Male)
8.1 (Female)
10.0 (Male)
5.8 (Female)12.5Secondary vs. Middle18526.3 (Pooled)
28.3 (Male)
1.9 (Female)22.0 (Male)
12.0 (Female)
26.0 (Male)
28.0 (Female)9.8 (Male)
15.0 (Female)7.4 (Male)
20.0 (Female)
12.0 (Male)
19.0 (Female)29.0University vs. Secondary3827.2 (Pooled)
29.3 (Male)
21.8 (Female)16.0 (Male)
3.6 (Female)
3.6 (Male)
28.0 (Female)25.0 (Male)
-0.1 (Female)21.0 (Male)
26.0 (Female)
13.0 (Male)
16.0 (Female)33.3SourceSiphambe
(2000)Amin & Awung
(2005)Schultz (2003)Schultz
(2003)Schultz
(2003)Lassibille
&Tan(2005)aThe comparison is between primary and none, lower secondary and primary, upper secondary and lowersecondary and between tertiary and upper secondary.bThe comparison is between primary and none, secondary and primary and between university andsecondary.cIn each row the first figures for male and female are for those within age category of 24 to 34 and thelatter figures are for those in the age group 35-54.dThe comparison is between middle and none, secondary and middle and between university andsecondary. In each row the first figures for male and female are for those within age category of 24 to 34
and the latter figures are for those in the age group 35-54.eThe comparison is between primary and none, technical and vocational education and primary, secondarygeneral education and technical and vocational education and between higher education and secondary
general education. Essays in Education Volume 21, Summer 200792 Conclusion, Implication and Recommendation Much research work on the private rates of return to education reveals the positive correlation between the amount of education an individual possesses and his/her earnings.
The familiar argument of the proponents of human capital theory is that people invest in
education and training so as to increase their future earnings. In his seminal work on the
subject, Mincer has estimated the relationship between education and personal income
distribution by developing an earnings function. However, the standard Mincer
regression, though it was a major breakthrough in labour economic research, has attracted
criticism from different scholars in the field. The standard earnings function (which uses
log earnings as dependent variable and years of schooling and years of work experience
as independent variables) was criticised because the model did not include other
productivity related variables, such as job tenure, type of occupation, sector of
employment and family background. The exclusion of such variables from the earnings
model, therefore, could create an upward bias. Besides, the basic earnings function is
inadequate to measure the private rates of return to education at different levels. Due to
these points, it is important to extend the basic human capital model. Using this estimation technique and based on the data collected the estimated results show that the private rate of return to investment in education in Eritrea is high.
From an individual perspective, the causal effect of education is large enough for
education to be beneficial. Not only every additional year of schooling causes a
significant rise in earnings but higher rates of return are found to be associated with
higher levels of education. Given that Eritrean workers do still benefit from large private
returns to their schooling, it may be argued that education is still a valuable investment
from the private point of view. Besides, the estimated private rate of return can be used to
explain the demand for education and assess the equity or poverty alleviation effects of
public education expenditures. An increase in years of work experience has also a
positive impact on earnings. The negative sign of the tenure coefficient implies the
comparative advantage of labour mobility. The statistically significant negative
coefficient on tenure signifies that the free movement of labour should not be restricted,
especially by public sector. Another alternative is rewarding tenure, thereby increasing
job satisfaction and labour productivity. The positive coefficient of the dummy variable
for gender implies that the wage differential between male and female employees is due
to gender or factors associated with gender that we have not controlled for in the
regression.11 There are two main policy implications of these findings. First, the large private economic return to schooling in Eritrea suggests the further need for public expenditure
on education. Second, the increasing pattern of private rate of return to education by level
of education also suggests that a shift of part of the education cost burden (specially in 11 Though it is beyond the scope of this study, male-female wage differentials can further be analysed usinga decomposition technique to estimate the portion of the wage gap due to discrimination and that part of the
gap due to differences in productivity related factors.Essays in Education Volume 21, Summer 200793tertiary level) from the government to the individual and/or his/her family could be done
through loan schemes. Therefore, there is an opportunity for private financing at the
tertiary level of education. As reported by the World Bank (2002), in year 2000, the total
spending per university student in Eritrea was 14.3 times as large as the spending per
student in elementary level, or 27.3 times the spending per student in middle level. In
2004, Eritrea’s public expenditure per student, as percentage of GDP per capita, was
9.8% in primary, 17.4% in secondary and 855.5% in tertiary level of education (World
Bank, 2006). This indicates that government spending per student is much higher in
tertiary level. However, if the government tries to share part of the total cost at tertiary
level (for instance by introducing a loan scheme), the outcome will be a cost effective
increase in stock of human capital. The high private rates of return to investment in education in Eritrea, but on the other hand the low stock of human capital, entail multifarious policy measures. Not only
there is a need for expanding access to education (by way of double or multiple shift at
school, multigrade classes, class size increase, and construction and rehabilitation of
schools) but also government subsidy or credit system should be introduced to motivate
people to attain higher levels of education. So as working children have a chance of
school enrolment evening classes or flexible school system should be introduced.12 For those who are not young, adult education is a necessary strategy for increasing their
levels of education and thus getting higher earnings. In Eritrea, the government’s strong desire to expand education, together with the adoption of long-term plan for expanding education, can be taken as encouraging factors.
However, the existence of high population growth, poverty and opportunity cost of
education is still a worrying factor.13 Therefore, besides policy on poverty and population and allocation of resources in favour of education, the government should promote
private schools, integrate different forms of education, expand early child development
programs, expand adult education, and let non-government organisations involve in
education. Simultaneously, education research program and follow-up and evaluation of
the education system need to be strengthened. Finally, by developing community
awareness of the advantage of education, it is possible to motivate parents to send their
children to school. The paper acknowledges that the exclusion of the self-employed wage from the analysis might create selectivity bias. For further work, therefore, it is required to model
and estimate an auxiliary equation that explains self-employment status using selectivity
correction techniques. 12 In Eritrea, the percentage of children under 14 working in the labour force is around 38% (World Bank,2004).13 While the annual population growth rate is around 2.3%, the percentage of population living under US$1a day is 69% (world Bank, 2004)Essays in Education Volume 21, Summer 200794 Acknowledgements I am grateful for the inputs, as comments and discussions of Richard Brown. References
Altonji, J., & Dunn, T. (1996). The effects of family characteristics on the Return to education. The Review of Economics and Statistics, 78 (4), 692-704.
Altonji, J., & Shakotko, R. (1987). Do wages rise with job seniority? Review of Economic Studies, 54 (179), 437-459.
Altonji, J., & Williams, N. (1997). Do wages rise with job seniority? A reassessment. Working Paper Series, National Bureau of Economic Research; inc., Working Paper 6010.
Amin, A., & Awung, W. (2005). Economic analysis of private returns to investment in education in Cameroon. Africa Institute for Economic Development and Planning (IDEP). Retrieved August 10, 2006, from <http://www.saga.cornell.edu/saga/educconf/amin.pdf> Armolaran, A. (2002). Private wage returns to schooling in Nigeria: 1996-1999. Discussion Paper No. 849, Economic Growth Center, Yale University, New Haven CT.
Arneberg, M. (1999). Feminization of urban Eritrean households: Implications for poverty reduction policy. Fafo Institute of Applied Social Sciences, Department of International Studies, Oslo, Norway. Retrieved August 10, 2006, from <http://www.fafo.no/ais/africa/eritrea/HouseholdFeminisationEritrea.pdf>
Beach, C., & Finne, R. (1988). Family background in an extended earnings-generation model: Further evidence. Eastern Economic Journal, 14 (1), 39-49.
Behrman, J., & Deolalikar, A. (1993). Unobserved household and community heterogeneity and the labour market impact on schooling: A case study for Indonesia. Economic Development and Cultural Change, 41 (3), 461-488.
Bennell, P. (1996). Rates of return to education: Does the conventional pattern prevail in Sub-Saharan Africa? World Development, 24 (1), 183-199.
Bigsten, A., Isaksson, A., Soderbom, M., Collier P., Zeufack, A., Dercon, S., et al. (2000). Rates of return on physical and human capital in Africa’s manufacturing sector. Economic Development and Cultural Change, 48 (4), 801-827. Essays in Education Volume 21, Summer 200795Bjorklund, A., & Kjellstrom C. (2002). Estimating the return to investments in education: How useful is the standard Mincer equation. Economics of Education Review, 21 (3), 195-210.
Dearden, L. (1998). Ability, families, education and earnings in Britain. Institute for Fiscal Studies Working Paper No. W98/14.
Duraisamy, P. (2002). Changes in returns to education in India, 1983-94: by gender, age-cohort and location. Economics of Education Review, 21 (6), 609-622.
Glewwe, P. (1991). Schooling, skills and the returns to government investment in Education. Working Paper No. 76, LSMS, Washington D.C.: The World Bank.
Griffin, P., & Edwards, A. (1993). Rates of return to education in Brazil: Do labour market conditions matter? Economics of Education Review, 12 (3), 245-255.
Gottesman, L. (2002). A survey of educational systems in Eritrea. In: World Education Encyclopedia: A Survey of Educational System Worldwide, ed. R. Marlow -Ferguson, Vol. 1, 2nd edition. pp. 410-419. Farmington Hills (USA), Gale Group.
GSE, Government of the State of Eritrea, (1994). Macro-Policy. Asmara: Eritrea
Halvorsen, R., & Palmquist, R. (1980). The interpretation of dummy variables in semilogarithmic equations. American Economic Review, 70 (3), 474-475.
Hartog, J. (1986). Earnings functions: Beyond human capital. Applied Economics, 18 (12), 1291-1309.
Hawley, J. (2004). Changing returns to education in times of prosperity and crisis, Thailand 1985-1998. Economics of Education Review, 23 (3), 273-286.
Jacobson, L., Lalonde, R., & Sullivan, D. (1993). Earnings losses of displaced workers. The American Economic Review, 83 (2), 685-709.
Jones, P. (2001). Are educated employees really more productive? Journal of Development Economics, 64 (1), 57-79.
Kazianga, H. (2004). Returns to schooling for wage earners in Burkina Faso: Evidence from the 1994 and 1998 national survey. Retrieved August 10, 2006, from <http://www.columbia.edu/~hk2252/Returns.pdf> Kothari, V. (1970). Disparities in relative earnings among different countries. The Economic Journal, 80 (319), 605-616. Essays in Education Volume 21, Summer 200796Krishnan, P. (1996). Family background, education and employment in urban Ethiopia. Oxford Bulletin of Economics and Statistics, 58 (1), 167-183.
Lam, D., & Schoeni, R. (1993). Effects of family background on earnings and returns to schooling: Evidence from Brazil. Journal of Political Economy, 101 (2), 710-740.
Lassibille, G., & Tan J. (2005). The returns to education in Rwanda. Journal of African Economics, 14 (1), 92-116.
Li, H. (2003). Economic transition and returns to education in China. Economics of Education Review, 22 (3), 317-328.
Lucas, R., & Stark, O. (1985). Motivations to remit: Evidence from Botswana. Journal of Political Economy, 93 (5), 901-918.
Mann, P., & Kapoor, B. (1988). Earnings differentials between public, private and joint sector in Punjab (India). The Journal of Development Studies, 25 (1), 97-111.
Martins, P., & Pereira, P. (2004). Does education reduce wage inequality? Quantile regression evidence from 16 countries. Labour Economics, 11 (3), 355-371.
Mincer, J. (1958). Investment in human capital and personal income distribution. The Journal of Political Economy, 66 (4), 281-302.
Mincer, J. (1970). The distribution of labour incomes: A survey with special reference to the human capital approach. The Journal of Economic Literature, 8 (1), 1-26.
Mincer, J. (1974). Schooling, experience, and earnings. New York: National Bureau of Economic Research.
Mirer, T. (1983). Economic statistics and econometrics. New York: Macmillan Publishing Co., Inc.
MOE, Ministry of Education, (1999). Education for all in Eritrea: Policies, strategies and prospects, Preliminary draft report. Asmara: MOE. Retrieved March 15, 2002, from <http://www2.unesco.org/wef/countryreports/eritrea/contents.html>
MOE (2000a). Enrolment data: Various Issues. Asmara: MOE.
MOE (2000b). Eritrea: Basic Education Statistics 1999/2000. Asmara: MOE.
MOE (2004). Eritrea: Basic Education Statistics 2001/2002. Asmara: MOE. Retrieved March 20, 2006, from <http:www.erimoe.gov.er/Pages/edustatistics.html> Essays in Education Volume 21, Summer 200797Moll, P. (1998). Primary schooling, cognitive skills and wages in South Africa. Economica, 65 (258), 263-284.
Moock, P., Patrinos, K., & Venkataraman, M. (2003). Education and earnings in a transition economy: The case of Vietnam. Economics of Education Review, 22 (5), 503-510.
Mwabu, G., & Schultz, T. (2000). Wage premium for education and location of South Africa workers, by gender and race. Economic Development and Cultural Change, 48 (2), 307-334.
Pritchett, L. (1999). Has education had a growth payoff in the MENA region? Middle East and North Africa, Working Paper Series No. 18. The World Bank.
Psacharopoulos, G. (1985). Returns to education: A further international update and implication. The Journal of Human Resources, 20 (4), 583-597.
Psacharopoulos, G. (1994). Returns to investment in education: A global update. World Development, 22 (9), 1325-1343.
Ransom, M. (1993). Seniority and Monopsony in the Academic Labour Market. American Economic Review, 83 (1), 221-233.
Rees, H., & Shah A. (1995). Public-private sector wage differential in the UK. The Manchester School of Economics and Social Studies, 64 (1), 52-68
Rena, R. (2003). Eritrean Education - Historical Perspective. Retrieved April 20, 2003, from <http://www.shaebia.org/artman/publish/article_641.htm> Sahn, E., & Alderman, H. (1988). The effect of human capital on wages and the determinants of labour supply in developing country. Journal of Development Economics, 29 (2), 157-183.
Schultz, T. (1999). Health and schooling investments in Africa. Journal of Economic Perspectives, 13 (3), 67-88.
Schultz, T. (2003). Evidence of returns to schooling in Africa from household surveys: Monitoring and restructuring the market for education.Discussion Paper No. 875, Economic Growth Center, Yale University, New Haven CT.
Shah, A. (1986). Education and earnings inequality across British labour markets. Applied Economics, 18 (2), 205-217. Essays in Education Volume 21, Summer 200798Silles, M. (2005). Personality, education and earnings. Centre for Applied Microeconometrics, Institute of Economics, University of Copenhagen. Retrieved August 10, 2006, from <http://www.econ.ku.dk/CAM/> Siphambe, H. (2000). Rates of return to education in Botswana. Economic of Education Review, 19 (3), 291-300.
Soderbom, A., Teal, F., Wambugu, A., & Kahyarara, G. (2004). The dynamics of returns to education in Kenyan and Tanzanian Manufacturing. CSAE Working Paper Series/2003-17. Stefanos, A. (1997). Women and Education in Eritrea: A Historical and Contemporary Analysis. Harvard Education Review, 67 (4), 1-26.
Terrell, K. (1989). An analysis of the wage structure in Guatemala city. The Journal of Developing Areas, 23, 405-424.
Theodossiou, I. (1996). Promotions, job seniority and product demand effects on earnings. Oxford Economic Papers, 48 (3), 456-472.
Topel, R. (1991). Specific capital, mobility and wages: Wages rise with job seniority. Journal of Political Economy, 99 (1), 145-176.
Trostel, P., Walker, I., & Woolley, P. (2002). Estimates of economic return to schooling for 28 countries. Labour Economics, 9 (1), 1-16.
UNIDO, United Nations Industrial Development Organization (2003). Eritrea: Human Resource for Sustainable Industrial Development. Retrieved April 20, 2006, from <http://www.unido.org/en/doc/16704>
van der Gaag, J., & Vijverberg, W. (1989). Wage determinants in Cote d’Ivoire: Experience, credentials and human capital. Economic Development and Cultural Change, 37 (2), 371-381.
Vieira, J. (1999). Return to education in Portugal. Labour Economics, 6 (4), 535-541.
Wambugu, A. (2001). Real wages and returns to education in Kenya manufacturing. Retrieved August 10, 2006, from <http://www.csae.ox.ac.uk/conferences/2001NIRaFBiA/pdfs/wambugu.pdf>
Williams, N. (1991). Re-examining the wage tenure and experience relationship. The Review of Economics and Statistics, 73 (3), 512-517.
Wooldridge, J. (2003). Introductory Economics: A modern Approach, 2e. USA: Thomson South-Western. Essays in Education Volume 21, Summer 200799
World Bank (2002). Eritrea: Education and Training sector Note. Report No. 24448-ERI, Human Development 1, Africa Region, Washington D.C.: The World Bank.
World Bank (2004). Africa Development Indicators 2004. Washington D.C.: The World Bank.
World Bank (2006). 06 World Development Indicators. Washington D.C.: The World Bank.
Yang, D. (2005). Determinants of schooling returns during transition: Evidence from Chinese cities. Journal of Comparative Economics, 33 (2), 244-264.

refer page:-------http://www.officesoon.com/doc/194262-the-private-rate-of-return-to-schooling-evidence-from-eritrea-temesgen